Do one thing right now, before you start reading. Open ChatGPT, Claude, Gemini, or Perplexity (whichever one you usually use) and ask, just as your client would: "Recommend the best companies in [your niche, your region]." Take a look at the list.
Are you not on the list? Or are you there, but only after three competitors and with an error in the description? The good news is that you're not the only one missing from the list so far. The bad news is that your competitors are already there.
Next, we'll break down why this is happening, who actually makes it onto these lists, and where your leverage lies. There's a lot of text because it's based on research from 84 sources, ranging from arXiv to Bitkom. But at the end, a visibility test awaits you: 10 search queries, a table, and an explanation. It's free, quick, and most importantly, it includes tips on what to do right now to turn things around.
Consumers are already asking the machine
Let's start with the scale of the phenomenon. To give you a better idea.
In Germany, half of all internet users already replace traditional search with AI chatbots at least occasionally, and among those under 30, the figure is two-thirds (Bitkom, 2025; telephone survey of 1,156 people, a representative sample with no marketing bias). In Austria, 70% of consumers use AI services at least occasionally, and among Gen Z, the figure is 95% (AI Readiness Study, 2025). Switzerland: 38% already use ChatGPT while shopping (AlixPartners, 2025).
And now for the financial aspect. Ketchum/YouGov (2026) surveyed those who already use AI: 46% consult it regarding purchases, rising to 64% among the 25-34 age group. Trust in AI recommendations is higher than in influencers: 18% versus 11%. When considering all consumers, the figures are more modest, but the trend is the same: 31% of Germans research products via AI chatbots, and a third of them do so more often than they use Google (ECC Köln, 2025). And this is no longer unique to Germany: globally, 58% of consumers cited generative AI as their primary source of recommendations over search engines, up from 25% two years earlier (Capgemini, 2025).
In the B2B sector, the situation is even more intense. According to Forrester, 94% of business buyers used AI during their most recent purchase, up from 89% a year earlier. Gartner cites a more modest 45% (we're including both figures because the truth likely lies somewhere in between): it depends on how the question is phrased. The most telling statistic: among software buyers, 51% start their search for a vendor with an AI chatbot and only turn to Google afterward. In April 2025, that figure was 29% (G2, 2026). That's a doubling in just one year.
And here's the result that prompted us to write this article in the first place: 69% of B2B buyers chose a vendor based on an AI recommendation, while 33% purchased from a company they had never heard of before (G2, 2026). A third of the deals went to the companies the AI suggested. If you weren't in the recommendation, you lost a bid you didn't even know about.
One caveat to keep in mind: people still make the final decisions themselves. 69% of B2B buyers verify AI insights with a live sales representative (Gartner, 2026), and only 11% are willing to let the machine handle the actual purchase. AI dominates the research and shortlisting phases. But when a buyer makes a call, they're already reaching out to those on their shortlist. If you're not on that list, there won't be a conversation.
Google Isn't What It Used to Be, Either
Even if your customer hasn't left Google, they've still ended up in a new layer. Google shows AI Overviews, generated answers displayed above search results, to 1.5 billion people every month in over 200 countries (Google, 2025). The impact on websites is clear: about 60% of searches end without a single click (Bain, 2025; Deloitte estimates 55-60% for Germany).
However, this traffic hasn't disappeared. It has simply shifted. In June 2025, AI platforms generated 1.13 billion visits to websites, a 357% increase from the previous year (Similarweb). And this traffic converts at a rate of 7.1%: the second-highest result after paid search at 7.8%, higher than organic search, social media, and email (Similarweb, 2026). The fastest-growing channel also drives the most engaged visitors. We spent a long time looking for a catch in this data. We didn't find one.
How the Machine Decides Who to Mention
Now for the main question: What principle does the AI use to compile its list?
The short answer: It listens to what others are saying about you.
Ahrefs analyzed 75,000 brands and identified which factors correlate with appearing in AI responses. Brand mentions across the web showed a correlation of about 0.66. Traditional backlinks and domain authority: only 0.27-0.33, half as strong. And the strongest single signal turned out to be mentions on YouTube, at 0.737. In our opinion, this is the most underrated channel in this whole story: site owners invest in their websites and forget that AI watches videos (or, more precisely, reads transcripts).
Academic data confirms this. Researchers at the University of Toronto conducted a series of controlled experiments on three AI search engines, in several languages, using rephrased queries. The result: a systematic bias in favor of earned media, that is, third-party authoritative sources. They cite the brand's own content and its social media posts much less frequently (Chen et al., 2025). 82% of all AI citations come from third-party platforms (Ahrefs, 2025).
The mechanics become clear if you look under the hood. Modern AI search engines cross-check sources with one another and weigh them by reliability before providing an answer; roughly speaking, they hold a vote (Hwang et al., EMNLP 2025). For a brand, this means one simple thing. A single mention, even a brilliant one, carries almost no weight. What matters is the chorus: several independent sources that all say roughly the same thing about you.
The "About Us" page isn't singing in this choir. It's sitting in the audience.
The list isn't neutral. And it fluctuates.
There are two things you should know about these lists in advance.
- First: the models have their favorites. GPT recommended Samsung in 97.1% of smartphone queries and Tesla in 92.2% of electric vehicle queries; American models rank U.S. brands at the top about 140 times more often than their European counterparts (Rienecker et al., 2026). That sounds unpleasant for a company based in Munich or Kyiv. However, there's one detail that changes everything: this bias is primarily detected by search engines with live web search, while a "pure" LLM relying on its own knowledge turns out to be more balanced than classic recommendation algorithms (Lichtenberg et al., 2024). To put it in layman's terms: the bias exists on the live web that the machine reads. And the web can be influenced.
- Second: the answers vary. None of the tested models ever produced identical answers across five repeated runs, even with "creativity" turned off (Atil et al., 2024). The set of cited sources matches between runs by at best half (Sielinski, 2026). Today you're cited; tomorrow you're left out. Keep that in mind. Later, when we get to the test, it will become clear why we're running each query twice.
GEO, AEO, and Why Your SEO Hasn't Made the Move Yet
Names have already been coined for this new discipline: GEO, Generative Engine Optimization, optimization for generative search engines. The term was coined by researchers at Princeton and Georgia Tech back in 2024, who demonstrated that the visibility of a source in AI-generated answers can be increased by 40% (Aggarwal et al., KDD 2024). Closely related is AEO, Answer Engine Optimization. The difference is explained in a single sentence: SEO means being on the list; AEO means being the answer.
The most important takeaway from that same study: traditional SEO tactics don't translate. Stuffing a page with keywords in generative search yields, to quote the study, "little to no improvement". However, adding statistics, expert quotes, and links to sources boosts visibility by 30-40%. The irony is that these are quality factors that the SEO community has for years dismissed as "just nice to have." Here, they've become the ticket to the top.
Do the Non-Giants Stand a Chance?
It's that whole "it's just Amazon and Wikipedia anyway" thing. But no... The data is more complex. And more encouraging.
Yes, there is concentration: a handful of the largest domains account for a disproportionate share of citations, and in e-commerce, Amazon appears in 61.3% of ChatGPT's responses (BrightEdge, 2025). But look at the long tail. 84% of all ChatGPT citations come from sources outside the top 100, provided by over 148,000 different domains (AirOps, 2026). In Google AI Overviews, only 38% of citations come from pages in the top 10 organic search results (Ahrefs, 2026): a high ranking in traditional search is no longer a guarantee of inclusion. The choice of search engine also matters: Perplexity cites over 8,000 different domains, compared to approximately 2,100 for ChatGPT (BrightEdge, 2025).
It gets even more interesting when it comes to optimization. In the Princeton experiment, a page ranked fifth saw a 115% increase in visibility after citations were added, while the then-leader dropped by 30% (Aggarwal et al., 2024). This new layer levels the playing field. You can figure out for yourself who stands to benefit.
And the composition of the top tier isn't set in stone (that phrase sounds familiar). Reddit's share of ChatGPT citations plummeted from about 60% to 10% in a few months, while Wikipedia's dropped from 55% to less than 20% (Semrush, 2025; sample of 230,000 prompts). Doors and windows (you name it) are regularly opening up.
That's our conclusion, and we're ready to defend it. Major brands win out on general queries like "recommend a CRM," because the model's memory of well-known names comes into play there. But a specific query ("CRM for dental clinics with 5-10 chairs in Bavaria") is handled through live search, where a niche specialist competes with other specialists just like himself. For a profitable company, whether it has a 1-year or a 15-year track record, it's much easier here than it seems after reading the first few paragraphs.
DACH Snapshot: SME Survey and Language Issues
In the spring of 2026, a well-known Austrian agency ran 150 medium-sized companies from Germany, Austria, and Switzerland across 11 industries through ChatGPT. The results were such that we first went to verify the source. The model did not mention 56% of the companies at all. In 96% of cases, it made up the CEO's name. Only 4 out of 150 companies were represented completely accurately (maxonline, 2026). However, a small caveat: this is a press release from a commercial GEO agency, without peer review. But the picture aligns with what we see in our own checks, so we're sharing it here.
A separate note on language. The training data for large language models is heavily skewed toward English: Common Crawl, one of the core corpora, consists of about 300 billion pages, predominantly in English (Nature, 2025). No one has yet published an exact figure showing by how much German-language content lags behind English-language content in terms of citations; we looked for one. But the structural bias is well-documented, and it compounds the brand bias.
And here's another figure that explains the value of each spot on the list. A classic search returned an average of 10 links. An AI-generated response cites 4.3 sources from 3.4 domains (arXiv, 2025). A shorter list means one simple thing: there's almost no "second page" where you can fall through the cracks and still survive. Either you're among the three or four named, or you're not there at all.
What Really Drives Visibility
The full recipe won't fit into an article, and promising to include it would be misleading. But we'll list the main levers, complete with numbers.
- First and most challenging: mentions on third-party platforms. We've already discussed these above. We'll just add that this is a long-term game: industry reviews, podcasts, YouTube, and professional communities. There are few quick wins here.
- Second, something you can implement as early as tomorrow: evidence within your content. Statistics, expert quotes, links to sources, that's the very 30-40% increase from the Princeton study. Check your key pages. How many specific numbers are there, and how many general terms?
- Third: freshness, and its importance surprised even us. Seer Interactive analyzed over 5,000 pages cited by AI: 89% had been updated within the last three years, and only 6% of the cited content was older than six years (Seer, 2025). Perplexity generally draws half of its citations from content published this year. That 2019 article you're still proud of might as well not exist to the machine.
- Fourth: reviews. A third of Google AI Overviews responses reference at least one review platform (SE Ranking, 2026), and Perplexity does this almost always for consumer topics (Feefo; vendor-provided figure, take it with a grain of salt). If customers love you in silence, the machine won't know about that love.
- We saved the fifth one for last because it's surrounded by the most myths: structured data. The Organization schema is currently being marketed as a magic button. The reality is different. In Google AI Overviews, it correlated with an increase in citations, but in ChatGPT and Copilot, the effect was even negative, and 6 out of 7 platforms failed to correctly parse JSON-LD directly (Otterly, 2026). We still recommend implementing it, but with the right expectations: it's like an ID so the machine doesn't confuse you with anyone else. IDs don't win contests. Without them, you simply won't be allowed to proceed.
How Much Does It Cost
In Germany, AI Overviews appear for roughly one in five search queries and account for about 265 million organic clicks each month; the click-through rate for the top position has dropped from 27% to 11% (Sistrix, 2026). Pew Research, which has no SEO interests, confirms this trend: when an AI answer is available, users are half as likely to click on regular results.
The market has responded with capital. GEO tools have already attracted over $200 million in venture capital; category leader Profound raised $96 million in February 2026 at a valuation of $1 billion and serves one in ten Fortune 500 companies. Incidentally, the fastest-growing European player is based in Berlin: Peec AI doubled its annual revenue to $10 million in just six months. Vienna-based Otterly offers monitoring starting at $29 per month. The AI visibility measurement industry is already well-established, and this is the best proof that the problem is real.
The German corporate mainstream is already here, too. Deloitte writes: AI visibility today determines whether brands are even found at all. In a BVIK industry survey, 86% of B2B communications professionals cited working with AI as essential within the next two years. Even Gartner, back in early 2024, projected a one-quarter drop in traditional search volume by 2026; Gartner itself cautioned against interpreting this as a prophecy, but it did get the trend right.
The math adds up, and it's not good news for those who procrastinate. The organic channel you've been building for years is shrinking every month. The new channel is growing by hundreds of percent, converts better than almost everything else, and is still inexpensive because most of your competitors aren't active there yet. Windows of opportunity this big close quickly.
Test Yourself. Today
All these statistics will remain abstract until you see your own numbers. To help with that, we've put together the "Does AI Recommend You?" test: 10 queries, a table, and an explanation. It's free and provides three things that the one-time query you ran at the beginning of the article won't give you:
- a real-world picture: 10 angles from which buyers are searching for you, ranging from a query with your USP (without a brand name) to the awkward "Who would you NOT recommend this to, and why?" and the control question "How do you know that?";
- diagnosis: each query is run twice across three search engines; the results are tallied automatically, and you see your visibility level, ranging from "invisible" to "visibility leader", with an explanation of what it means for your business;
- the first three steps for this week: what to fix, what to add, and what to check next.
What to do with the results? For example, if the picture turns out to be grim and you want to dig deeper, we offer the comprehensive RAD360° diagnostic: AI visibility is just one aspect of it, but we look at the entire system that determines whether or not customers find their way to you. We'll explain, demonstrate, and then even integrate everything needed to ensure you're featured in future results. And if you're already among the AI's recommendations, then... congratulations! We'll invite you to share your success story with our readers and viewers.
Ready to find out if the AI recommends you when a buyer asks, "Who should I buy from?"
Buyers are already asking. The AI is already answering. All that's left is to find out whose name it mentions. Your move.
